Much like weather forecast models predict the trajectory of a hurricane, computational tools known as molecular dynamics simulations predict the microscopic motion of an interacting system of molecules. Molecular dynamics has sometimes been referred to as a “computational microscope” because it allows scientists to virtually “see” what is going on at the nanometer scale, the length scale of proteins and biological macromolecules. Even with significant advances in computer hardware, however, molecular dynamics are simply not long enough to simulate many significant biological processes like protein interactions.

Part of the reason that the simulations are slow is because they spend a lot of time watching the same bonds vibrate back and forth. These motions, however, are not vital to understanding the system. If we want to predict where a fishing boat is moving, we probably don’t really need to know how the fisherman’s hair is waving in the wind. The most relevant information to our prediction is the position of the oars, how fast they are pushing against the water in the lake, and the direction and size of the waves. In constrained molecular dynamics, certain motions are suppressed, allowing simulations to proceed more quickly.

Unfortunately, constrained molecular dynamics disturbs the population of microscopic states. If standard unconstrained molecular dynamics gives you fair dice, constrained molecular dynamics gives dice that are crooked. Laurentiu Spiridon, former senior research associate, and David Minh, assistant professor of chemistry, developed and applied a technique to make the crooked dice fair again, allowing those interested in simulating molecular motions to use the time-saving method of constrained molecular dynamics without messing up their final predictions.

Spiridon and Minh applied their method to a number of simple model systems to show that they did indeed implement the correction properly. They also demonstrated it on a set of macrocycles — organic molecules that contain a large internal ring — which are challenging to sample with traditional molecular dynamics. In the latter calculations, they were able to observe a much larger number of conformations in the same amount of computer time. They anticipate that the method will be particularly useful for modeling large-scale conformational changes in modular proteins.

Their paper describing these results, “Hamiltonian Monte Carlo with Constrained Molecular Dynamics as Gibbs Sampling,” has been accepted in the Journal of Chemical Theory and Computation and is available here. In 2016, this American Chemical Society journal had an impact factor of 5.245, top among journals specializing in theoretical and computational chemistry.